***Please note that this is an automatically generated transcript and will contain errors.***

---

So we're going to talk about geo decentralization today and why it is a foundational question. I think, you know, so what is this? This is like, I just said earlier on the baked work, so it takes me, on average, like 5 to 10 years to write a definition paper. Even though the definitions usually have like approximately 20 characters.

So it's about a character every six months. So if you're complaining that this isn't progressing fast enough or whatever, I'm sorry if that's just how I work. And so this is going to be some, oops, some napkin, research, some kind of sketches of some in progress stuff. For a lot of you, all this may have some old content, from this told me to assume people are not that familiar with, life building and Ethereum, which was not the crowd.

I'm not sure if that was an accurate representation. So I'm sorry about if it's a little boring, but I will try to say some spicy new things. You can get more expert. So today we're going to motivate the need for a new science. So that's the foundational part is the new science. For internet scale distributed systems such as Ethereum, we're going to explain the importance of global systems.

I actually might have forgotten to do that part. So we might not do that. But everyone here is in crypto and you've been talking about it. So we can assume global systems are important. And then we're going to tie in the intersection of economics and distributed systems. So when blockchains came about there was this new interest in this field of study called crypto economics.

So before let's say you were deploying, I don't know, cloud app for Facebook. You could just choose your five files have a metric nature. They were sufficiently geographically distributed. And because you controlled a lot of the substrate, that would be good enough. You wouldn't have to risk those nodes moving or those nodes kind of doing all these bizarre behaviors, or those nodes trying to profiteer on top of your system, or all of these things that we see in crypto systems.

On the other hand, with crypto systems, we aim for a much more kind of self-regulating ideal, a more market based, notion of security, and a more permissionless system where anyone can join or leave, or indeed move at any time. So how to secure systems in this model? That's kind of a new problem that before cryptocurrency didn't have much motivation.

And so I'm going to argue here that we need to go beyond that and kind of bring geography and almost physics into crypto economics, by analyzing a few simple examples and using them to make some pretty nuanced argument. So this is going to be a new science. There's a lot of work to be done. This is kind of, again, just my notepad of things that I've been thinking about on this and why I think it's important.

So these are my personal axioms on decentralization. I have been in crypto for way too long at this point. And the reason for that is because I care about these axioms, so you may not share them. That's fine. I don't really care, but they are my axioms. And so my axioms are as follows. The first one is the only decentralization that matters is that of power.
So things like that, counting cloud regions or service providers, if anything, very loose decentralization of power. What I'm talking about is things like decentralization of rewards, economic upside, arbitrage, financial activity, technical power, etc.. So in that vein, architectural or technical decentralization as it's typically studied and consensus and BFT will arise if the power is decentralized. So we know how to run a decentralized protocol on top of a decentralized substrate of power.

But we do not know necessarily how to incubate from scratch this global, decentralized, system of power, on top of the technical substrate in this permissionless, kind of self-regulating model. So that is what this foundational science aims to kind of illuminate. How do we actually do that? So therefore it should be a goal of all of these technical decentralization inquiries.

So when we're doing things like simulating or measuring decentralization, as we just saw or proposing protocols, as we're going to see later, the goal of these technical efforts should be to serve and protect the global decentralization of power that cryptocurrency intends to build. That's kind of the why, of all of this and the last axiom I have is that decentralization of power must be global.

If it's not global, it's fundamentally not as decentralized as it could be. Then I view we have failed as a community, and I will give cryptocurrency and go blow glass in the mountains. And that's about it. So these are my axioms. And so we're going to talk about today what we need to actually say from a technical perspective to decentralize power.

This was a slide I gave at SBC two years ago where I said, if you could solve this, my view is you would deserve a Nobel Prize in economics. And because there is no Nobel Prize in consensus, if you're a consensus feature, this is your only path to the Nobel Prize making for with names. Below you have a think that qualifies for a Nobel Peace Prize.

If you can decentralize power, you know, we've invited. You might be the first one to get a toast in the same year. I actually don't know if that's happened, but I'm guessing that hasn't. And you'd be the first, so, let's do it. Okay. So how did decentralize power.

Okay, so I'm going to basically spend this talk talking about one very, very microcosm question of this big field of study, that I'm trying to establish, which is that to decentralize power, the first step is to understand how not to pour fuel on the fire, of centralizing it more, if you can't figure out how not to make things centralized more, how can you expect to decentralize them? Right. This seems kind of obvious. And the corollary of this is that the world is already pretty centralized. Things like fiber roots, here, net things like financial activity, things like exchange locations, where to infrastructure extremely centralized. So we already are building on top of a very centralized substrate trying to build decentralization.

And how do we do that? Okay. So let's start with kind of a didactic example here, a very simple example. So Alice and Bob have to sign a transaction together. Hopefully all these slides make sense as I made them in like five minutes this morning on eight red balls. So people will roll with this. They don't.

So Alice and Bob have to sign a transaction together. And to be valuable, this transaction has to make it to the next block, let's say some bonding to some liquidation or something like that. Through a multi-sig that both of these parties control. And, so let's instantiate two protocols. A natural thing to do as technologists would be to throw technology at the problem.

So what technologies do we have? There's two that are kind of cool here. Among many I have a big table that's not in the slide deck, if you care later, but, MPC and these are two technical protocol things you can do. So in MPC, maybe Alice and Bob used the garbled circuits protocol to compute this joint signature on top of each of their private inputs, and then send it into the block, which maybe gets it via a smart contract.

The second possibilities, we can use a TDE to do this. So, Bob can send an authorization to Alice's TDE, which can produce a joint signature to be validated by the blockchain. Only after checking bob infestations. These are two choices for how to actually do this. How do we decide which one is going to be more decentralized?

How do we also parameterize these two protocols? Choose names like The Matrix, and various other things you need to tell them when you actually pull them. So I'm going to start by presenting this definition without context. Many of you have probably already seen this. It is basically a ratio. Or you can do a difference if you want to use the additive definition, which is probably better between the profit that a player and arbitrary rational player.

Again, we're trying to build cells. So hearing systems will make in a game and some network. So it's a ratio of the profit that this player will make. If everyone has no latency in the network to the profit that this player will make in a real world network. So if this ratio is very high, this means that this arbitrary player key is going to make a lot, a lot more when everyone is co-located than they would if they were running the same protocol on the previous slide across a worldwide network, where something like the internet, and we call this the geo centralization coefficient of the protocol, it's not trying to measure how centralized or decentralized you are or how many data centers you deployed in or anything like that. It's more in the vein of what point is set where we're trying to study. Basically the pressure on your protocol, what is the iterative pressure to centralize and to move closer together? In my view, in economics, the iterative kind of marginal pressure to act in a rational model determines a lot of the outcomes.

And certainly a lot of the moments, we see in economics. So that's why we study it this way. So how do we actually study these two protocols? Okay. Well, let's say that we instantiate this concretely into a game where Alice and Bob have five seconds and we will make $100 if they can respond to this observation.

So let's consider two different, we're also adding some stuff to the protocol specifications up top. So in MPC, we know from verbal circuits that Alice and Bob have to exchange a lot of messages over whatever communication channel they have to divide an output. We have to kind of exchange messages corresponding to logical gates. So let's say this takes around 20 messages to generate this signature securely with a TV.

This would require, let's say three messages, one from Bob to Alice, one from Alice to the chain. I don't know why it's three and not to whatever. It still works. So let's compare these two protocols in a game where they have five seconds and gain $100 for completing this kind of task when the network delay is something like .01 seconds in the system.

Let's say that's the worst case delay that could ever be expected on the communication channels between Alice and Bob. It doesn't really matter. Neither one of these schemes will yield Alice and Bob more profit from Co-locating, because their worst case latency is already sufficient to gain the full reward. So in this case, both schemes make $100.
If, on the other hand, they have a three second network latency, obviously these 20 messages can't happen within the deadline. So that scheme will make no money. These three messages, on the other hand, that should be one second, not three seconds. Sorry, but, you get the point. To invest money. Yeah. Like I said, it was a long morning.

So in this world, SGX protocols for this game actually can be more geof centralized than ABC because they can be deployed across slower communication channels and yield new, the same time profit curves that you would see through something like that RPC protocol. Very counterintuitive for crypto people. All right. Let's bring in blockchains because we love blockchains. This is where if you've seen all this before, I might say something that's actually interesting to you.

So turn on here. So let's consider two blockchains, both in let's say the stream or Nakamoto style for choice consensus rules. Let's not complicate things on the left. We're going to prioritize the blockchain terminal second block time. And on the right within a friend or Isaac with a five minute block time. And let's say that most of your validators in this example are located in a single data center, let's say 60%.

We see this concentration in Aetherium and things like us East in Solana, in places like hedge, etc.. And so let's look at, let's choose this arbitrary key that we were looking at in our definition earlier to be this player that's kind of on the outskirts of the network and has 20 milliseconds of latency to this cluster.

So now if we compare basically how much profit they would make if they were co-located, how much they make in the real world on the left side, where they have ten milliseconds of block time, this player makes no money at all because they're just not able to send any messages or respond to the chain. The blocks are progressing faster than they can even send a single message.

So there's really no health care for this player. And on this, on this side, oh, you get the call. The teddy bear or the really angry. Very, those are from old slides, on this side. Happy face. And still happy. Because they can contribute. They have 20 milliseconds. They can run and upload a protocol. Now, the interesting thing I was going to say is talking about proof of location.

You can actually invert this property to look at proof of location. So what you have on the left side is both a very geo centralizing consensus protocol for something like Etherium layer one. But if you wanted to use something like a local currency, how perfect is something like this be? If you already knew a lot of your, economy and a lot of your power was in one location, it is incentivizes people to be increasingly far from that location, through a very physical, kind of technical, substrate constraint.

And so I think we could actually use this to build things like space beacons, which you'll hear about later. But even simpler than that, just local currencies that run out of, physically centralized points. And that might be an interesting way to both get water flowing decentralized further. So this is bad for decentralized layer one. So it's not necessarily bad for all systems.

It depends really on what you're trying to do and what market you're trying to serve. Are you trying to do global or not? All right. I'll probably wait for time. That's okay. No I'm not. Well, So, okay. So, rainbow staking, this is a big trend in Ethereum, which I think, kind of demands a lot of geo of centralization thinking.

So this is the idea that before where you would have one class of state validators securing the network, and these validators were kind of fungible against how much, ether they stake in the protocol. Now, we have designs being considered where these validators are broken into classes that they can sign up for based on their technical capabilities.
And these classes may actually require different, latency profiles, different hardware requirements, different, etc., etc.. You've seen the acceleration of this with things like Solana through like the new gateway based consensus protocols are using their etc..

And so in this world, what becomes important is actually looking at each of these classes and thinking about as a network, how decentralized do we actually want each of these things to be?
How do they affect each other? If one of these things is extremely centralized, so does it affect the rest of the network or not? And I think this is in a conversation. We've had a lot in here. Specifically two interesting problems that we can care about. One is block building. We've talked about it before today. How to know what we actually put in a block of the blockchain.

How do we resolve conflicts between different conflicting transactions? How do we assign arbitrage profits for latent state on this global system? These are the key questions when it comes to what goes into the blockchain. Right. And so if you're going to handle this to a class of stakers, how decentralized do these stakers need to be? And how is that different from, let's say, something like attestation, which we just got a big simulation done, but I'm not gonna, introduce that.

So how do we actually compute what matters here, which is the time profit curve? Given the network distribution, and I don't have an answer for you. Again, these are these are notes of what I think is valuable to answer. And I'm trying to build out like a didactic model for the fundamentals here. But I have some sketches.

I'll show them in a second. I've also added some information here, which I think is is interesting and valuable. So if you're thinking about modeling these things, for example, building is a winner, takes all money. Market is an option with a single winner. Winner takes all latency sensitive, right. That is very different than the attestation market, where it's a market where kind of everyone who is contributing is splitting a reward evenly.
There are still tiny gains within the market who just talked about certain testers dropping off, but they're much, much less severe because of certain properties of how these rewards are actually distributed within the system. So those are very, very key things to both study and included in your models. So my, my, my, my sketches for what these time profit curves might actually look like in at least some worlds, in interesting and meaningfully different ways.
So looking at let's look at attestation first, bottom up, because it's a little bit easier in some ways. So an attestation basically you're going to get your share of the attestation rewards for that slot. If you submit a timely attestation to kind of be aggregated, and otherwise you're going to get zero. So here we have two worlds, one in which let's say the worst case latency and the latency of arbitrary p we're looking at is two seconds and one in which is five.
Between these worlds, the curves look very similar. You have a kind of a narrower region here in which this tester is able to respond, maybe if the proposed grade is playing timing games, this could even be different and shrink to zero. That's its own question. But still between, let's say, like a small, tiny differences. You have roughly similar looking, profit time curves.

That's at least my hypothesis. This is not proven. There's no proof of this. This is something that I would like to model, and I would like people here to help me model. On the talk. On the other hand, you have block building, which again, is this timing gains you winner takes all market. So let's say you have two blocks competing on the left one having a two second latency and on the right one having a five second latency.

Where they're taking these, five transactions that are arriving at random times across the slot, computing the value of each ordering of these transactions, and selecting and relaying the most valuable one, and whoever does that most efficiently kind of wins the auction. Well, on the left we see the guy who has two seconds or the user who has two seconds.

They have this kind of linear curve. So the longer they can wait into the slot, the more they're going to make because they can optimize this further. This will drop to zero at some point. I didn't draw it. Meanwhile, if you're the second place block, you're operating, with a five second network delay, you may kind of only get rewards if you are lucky in the entropy that you happen to try, in order that the person with the timing advantage hasn't already had plenty of time to explore and optimize further.

So you get kind of some small share of the blocks, but it will hover around a very low percentage. And in fact, that's exactly what we see in the market from, with, higher latencies. So how do we actually model these things from this foundational building block theory perspective? And what do we actually measure? Well, two interesting things to me.
One is like if we can actually draw these curves for very simplified and didactic versions of popular protocols, how, robust are these curves to things like changes in assumption and parameter? So if we move the network away from 2 seconds to 5 seconds, how similar are these distributions of profit? For those players, for those validators. And if we can't model this theoretically, maybe this is one thing to ask the simulator.

And another one is like how well distributed are expected rewards with heterogeneous players. So if you have things like the two second player and the five second player, how similar are at first, in equilibrium? I think those are the things that actually determine network decentralization. And so for rainbow, staking, every kind of rainbow class needs its own, geo decentralization goals and its own kind of studies, analysis.

With regards to these formulas, we should be able to profit. Is it only us? One? Is it only us one and only H plus some proxies? Is it everybody? Is it something in between? I don't actually have answers. I think this is a political question. Right. It may be valid to say like, hey, if we want everyone to participate, we can only get five second blocks.
And really, if we have this one part of the protocol centralized over, should we get two milliseconds and maybe that's more desirable, I don't know. Again, my axiomatic view as like the kind of, evil globalist and, crypto maxi is that we we do need the maximum level of decentralization for the systems where it matters most, like ESL one.

So let's build that. Let's not lose hope, and let's stop studying Byzantine fault tolerance and things like technical kind of assortments into buckets of like where is in which pocket is what after. And let's start really looking at like granular incentives of network convergences over time, for these new types of systems that we're building in the crypto, economic and crypto context.

All right. Thank you. That's all I have. I only got time for question seconds.

I have to get the whiteboard anyways. You might as well. All right then. Anyway, the space given. What was I space? Peter, actually. Wait. You guys. Yeah. Yeah, yeah, yeah. So I mean, like like, imagine if you had a geosynchronous satellite, right? That you could, you could measure as geosynchronous, then you could write a really fast blockchain off of that satellite.

And then the only people who would be able to sync and meaningfully, let's say, transact or operate infrastructure within that would be in like this digital phone of like leads to that like very extreme parameter for this region. Like, you know, it's all for the flashlight volumes right now. I don't know if that if that makes sense.

Yeah, I could watch just how those follow this model path in the context of Mars, in the context of Mars. Yeah. I think that's a very interesting unsolved question that I haven't fully thought about. I, I know I, I've heard rumors that like, apparently Elon has become, like, not Mars killer anymore because he became convinced, like the wolf eyes would follow him to Mars and, like, destroy all of his efforts to create his, like, greeting, you thanks to Sophia or whatever.

So I don't know really what blockchains will like. Follow us to Mars. Maybe we'll have, like, go to Mars local and like Mars, Earth, soft straits. It probably means if we want, like, fair balance, cooperation between Mars and Earth, where no, neither one is dominant in, like, the political, game theory, we probably need a lot of times that are long enough to, like, stand to, and that inherently limits the collaboration that will be possible.
Like on that catastrophic great between Earth and Mars. And in some ways, also this incentivizes us from going farther and farther into space. So we probably need to send it off the counter and send it to that as well as. Excuse me. Yeah. I mean, I was just glad met because, you know, it's like speed of light.

As long as I'm fast and light, there's always so fast. Like there's a lot of Mars. Yeah, exactly. Yeah. So we'll need, like, we'll need like, like day long blocks or something. So really run the whole block building process like back and forth. Yeah. It'll be pretty, pretty slow. So maybe the Bitcoiners are right. Actually, about ten minutes.

It's fast. So. Yeah. I mean, what happens when you start getting to, like, Neptune or something that looks like. That's fair. Yeah. I have a Bitcoin. It's bigger with regards to the solar system than the town of Aetherium. With regards to the solar system. Yeah. So okay, so, that's highly enough. Hey, you you ready to. All right.

Thanks for entertaining me, folks.